My research interests are in the development and application of high performance linear algebra algorithms. Currently I am investigating the use of techniques based on singular value decomposition and eigenvalue analysis to applications in information retrieval, web link analysis, and clustering. At NERSC I was also leading the project on "Sparse Linear Algebra Algorithms for MPPs", which resulted in scalable implementation of the Lanczos algorithm, approximate inverse preconditioners, and dynamic partitioning. My continued areas of interest are sparse matrix algorithms, algorithms for large scale eigenvalue problems, and domain decomposition algorithms for unstructured domains for parallel processing. My algorithm research efforts were honored with the 1988 Gordon Bell Prize for parallel processing research. Software based on my "recursive spectral bisection algorithm" has been widely distributed to the finite element research community, and is regarded a major breakthrough in parallel algorithms for unstructured computations.

A second area of interest is in performance analysis of high performance computing architectures, and the assessment of supercomputing technology. I have been involved with the TOP500 effort almost since its beginning. I was member of the NAS team which developed the NAS Parallel Benchmarks, which has become an accepted standard for performance evaluation of MPP systems. I have been invited speaker at a large number of national and international conferences. I have given tutorials on high performance computing at many of the Supercomputing XX conferences as well as on many other occasions.

Specialties: - high performance computing, center management, performance analysis, benchmarking, architecture evaluation
- management of large R&D organization
- government contracts
- research in algorithms and applications on high end computing platforms